Predictors of hepatic encephalopathy in cirrhotic patients undergoing elective transjugular intrahepatic portosystemic shunt

Background: Transjugular Intrahepatic Portosystemic Shunt (TIPS) is used for the management of patients with complications related to portal hypertension. A significant post-TIPS complication is new or worsening hepatic encephalopathy (HE), occurring in up to 35% of patients; thus, appropriate patient selection is critical. Previous studies suggest age, prior HE, Child-Pugh (CPS) and MELD scores to be predictive of post-TIPS HE. The purpose of this study was to identify significant predictors of HE development in patients undergoing elective TIPS procedure and develop a model to risk stratify potential candidates. Methods: Cirrhotic patients who underwent elective TIPS placement between 1/2004-6/2016 with at least 1-year follow-up were included, yielding 95 patients. Relevant data collected retrospectively included: patient demographics, TIPS indication, etiology of cirrhosis, albumin, platelets, creatinine, bilirubin, INR, sodium, pre-TIPS MELD, MELD-Na, CPS, pre-and post-TIPS portosystemic gradient, pre-TIPS benzodiazepine or opiate use (BZDO), pre-TIPS HE occurrence, and post-TIPS HE frequency, as well as severity by West Haven criteria. T-test analysis was used for continuous variables and Chi-square tests for categorical variables, while multiple logistic regression model was performed using stepwise variable selection. Results: 95 elective TIPS were performed in 38 females and 57 males with a mean age of 55. HE occurred in 39% (37/95) of patients within 1-year post-procedure, 9% (9/95) of whom had clinically severe HE (defined as grade III or IV). Univariate analysis showed albumin \u3c2.5 (p\u3c.004, OR: 0.35, CI: 0.17-0.75), MELD \u3e13 (p\u3c.004, OR: 1.16, CI: 1.04-1.29), MELDNa \u3e 17 (p\u3c.015, OR: 1.11, CI: 1.02-1.20), Child-Pugh Score \u3e10 (p\u3c.001, OR: 1.89, CI: 1.31-2.71) and pre-TIPS BZDO use (OR: 2.57, CI: 1.07-6.16) to be significant predictors of HE occurrence within 1-year. Multivariate analysis showed age (p\u3c.04, OR: 1.06, CI: 1.00-1.21), MELD (p\u3c.04, OR: 1.13, CI: 1.01-1.27) Child-Pugh (p\u3c.003, OR: 1.81, CI: 1.23-2.67), and Pre-TIPS BZDO use (p\u3c.01, OR: 4.16, CI: 1.41-12.28) to be predictive for development of HE post-TIPS. These four variables were used to create a predictive model with AUC found to be .81 (.73-.90). Conclusion: Our analysis was able to refine data from prior studies by establishing predictive cutoffs for MELD (13), MELD-Na (17) and CPS (10) for the development of post TIPS HE. We also found that pre-TIPS benzodiazepine/opiate use was a novel and strong predictor of post TIPS HE. Finally, we were able to create a model from this data using age, MELD, CPS and pre-TIPS BZDO use that reliably risk stratifies patients for the development of post TIPS HE